diff --git a/my/core/_deprecated/dataset.py b/my/core/_deprecated/dataset.py
new file mode 100644
index 0000000..9cca2fd
--- /dev/null
+++ b/my/core/_deprecated/dataset.py
@@ -0,0 +1,12 @@
+from ..common import PathIsh
+from ..sqlite import sqlite_connect_immutable
+
+
+def connect_readonly(db: PathIsh):
+    import dataset  # type: ignore
+
+    # see https://github.com/pudo/dataset/issues/136#issuecomment-128693122
+    # todo not sure if mode=ro has any benefit, but it doesn't work on read-only filesystems
+    # maybe it should autodetect readonly filesystems and apply this? not sure
+    creator = lambda: sqlite_connect_immutable(db)
+    return dataset.connect('sqlite:///', engine_kwargs={'creator': creator})
diff --git a/my/core/dataset.py b/my/core/dataset.py
index 31de4f4..40237a0 100644
--- a/my/core/dataset.py
+++ b/my/core/dataset.py
@@ -1,31 +1,5 @@
-from __future__ import annotations
-from .common import assert_subpackage; assert_subpackage(__name__)
+from . import warnings
 
-from .common import PathIsh
-from .sqlite import sqlite_connect_immutable
+warnings.high(f"{__name__} is deprecated, please use dataset directly if you need or switch to my.core.sqlite")
 
-## sadly dataset doesn't have any type definitions
-from typing import Iterable, Iterator, Dict, Optional, Any, Protocol
-from contextlib import AbstractContextManager
-
-
-# NOTE: may not be true in general, but will be in the vast majority of cases
-row_type_T = Dict[str, Any]
-
-
-class TableT(Iterable, Protocol):
-    def find(self, *, order_by: Optional[str]=None) -> Iterator[row_type_T]: ...
-
-
-class DatabaseT(AbstractContextManager['DatabaseT'], Protocol):
-    def __getitem__(self, table: str) -> TableT: ...
-##
-
-# TODO wonder if also need to open without WAL.. test this on read-only directory/db file
-def connect_readonly(db: PathIsh) -> DatabaseT:
-    import dataset # type: ignore
-    # see https://github.com/pudo/dataset/issues/136#issuecomment-128693122
-    # todo not sure if mode=ro has any benefit, but it doesn't work on read-only filesystems
-    # maybe it should autodetect readonly filesystems and apply this? not sure
-    creator = lambda: sqlite_connect_immutable(db)
-    return dataset.connect('sqlite:///', engine_kwargs={'creator': creator})
+from ._deprecated.dataset import *
diff --git a/my/core/serialize.py b/my/core/serialize.py
index 563e114..b5b1b3a 100644
--- a/my/core/serialize.py
+++ b/my/core/serialize.py
@@ -2,7 +2,7 @@ import datetime
 from dataclasses import is_dataclass, asdict
 from pathlib import Path
 from decimal import Decimal
-from typing import Any, Optional, Callable, NamedTuple, Protocol
+from typing import Any, Optional, Callable, NamedTuple
 from functools import lru_cache
 
 from .common import is_namedtuple
@@ -250,7 +250,6 @@ def test_dumps(factory: str) -> None:
 @parametrize('factory', ['orjson', 'simplejson'])
 def test_dumps_namedtuple(factory: str) -> None:
     import json as json_builtin  # dont cause possible conflicts with module code
-    import orjson  # import to make sure this is installed
 
     class _A(NamedTuple):
         x: int
diff --git a/tox.ini b/tox.ini
index 0b75b44..02acdfc 100644
--- a/tox.ini
+++ b/tox.ini
@@ -43,9 +43,6 @@ commands =
         # tbh not sure if it even belongs to core, maybe move somewhere else..
         # same with pandas?
         --ignore my/core/orgmode.py \
-        # causes error during test collection on 3.8
-        # dataset is deprecated anyway so whatever
-        --ignore my/core/dataset.py \
         {posargs}